Review by Shreya Laddha, 2022 (BTech)
Course Offered In 2021- Autumn Semester
Instructors Prof. Preeti Rao
Prerequisites Few concepts from signals and systems are required. Should know at least one of Python, MATLAB or Scilab for programming assignments.
Difficulty Moderate (3/5). If you are regular in class and complete all the assignments, the course content is easy to follow.
Course Content Following topics are covered in the course- Speech production (physiology) and acoustic phonetics, hearing and speech perception. Speech analysis: time and frequency domain techniques for pitch and formant estimation, cepstral and LPC analysis. Speech synthesis: articulatory, formant, and LPC synthesis, voice response and text-to-speech systems. Applications: data compression, vocoders, speech enhancement, speech recognition, speaker recognition, aids for the speech and hearing impairments.
Feedback on Lectures All material is covered in detail in lectures but they are a bit fast paced. You cannot stop paying attention at any point in the lecture. We had the advantage of listening lectures later on (multiple times may I add), owing to online sem. The professor entertains all doubts and is open to discussions. If you attend lectures with attention and ask doubts, then this is an easy course.
Feedback on Evaluations Evaluation Scheme - Computing assignments (vowel synthesis, speech analysis, speech compression, word recognition) -> 30% Independent project, or paper review & presentation -> 15% Quizzes -> 15% Exams (midsem, endsem) -> 40% We had 3 computing assignments and 1 project along with two quizzes, midsem and endsem. The course is very heavy in terms of academic load. The assignments are hard and time consuming but they are fun to do. You need to complete these assignments as these help in understanding many concepts from the lectures and also come in handy during exams. All the written exams tested the understanding of the concepts taught in class. So, if you have paid attention in class and your concepts are clear, then you can score good enough. There is not much need of mugging for these.
Study Material and References Lecture Notes are sufficient for this course. The following reference books can be used for further reading - Douglas O’Shaughnessy, Speech Communications: Human and Machine, Universities Press (India) Ltd.,2001 Rabiner and Schafer, Digital Processing of Speech Signals
Follow-up Courses Should definitely take ASR (CS 753)
Final Takeaways The course load is heavy. You need to devote time to this course and be regular (and attentive!!) in lectures. You need to score good on assignments as they make up 30%. Just be clear on the concepts and you are good to go for exams. But overall, this is a genuinely good course. You will end up learning so many new things about speech that you can use in everyday life!
Grading Statistics:
